Chasing NFL Playoff Perfection (Part 2)

Perfection flirted back last weekend. Six games, six chances to get humbled by the NFL’s favorite pastime, late-game chaos. I walked out 5–1, which in playoff terms feels less like a record and more like surviving a minefield with one shoe missing. The lone blemish still stings. Green Bay looked home free for 53 minutes, exactly the kind of pick that feels smart until football reminds you it does not care about logic. Then, the Bears did what they have kept on doing this season, erasing a deficit in the final two minutes and turning a near-perfect card into a reminder that no lead is ever safe and no prediction is ever finished.

Still, that is the hook. You do not chase perfection because it is unrealistic. You chase it because it is fragile. One snap, one read, one bounce of an oblong ball can undo hours of careful reasoning. And yet, here we are again. The divisional round raises the stakes and tightens the margins. Fewer teams, sharper strengths, smaller mistakes with bigger consequences. The noise fades a bit, the contenders reveal themselves, and the models get tested against teams that actually know how to win in January. I am back, not pretending the chaos is gone, but trying to see how long I can stay close to flawless before the NFL inevitably pulls the rug again. This is Chasing NFL Playoff Perfection, Part 2.

For my previous predictions, visit Wildcard Weekend

 

Divisional Round Weekend

Data

Every good prediction still starts with good data, and nothing about that changed from Part 1. I pulled the same numbers from the same trusted corners of football analytics: pro-footballreference.com and nfelo.com. PFR provided the raw production, while NFelo added the context. The goal, again, was balance.

That balance matters even more here. With only four seasons of seven-team playoffs, the divisional round gives us just 16 total games of historical data. That is nothing in NFL terms, and variance looms over every conclusion. Counting stats can lie, especially in small samples, where bounces, field position, and game script do a lot of the talking. By combining surface-level production with deeper efficiency metrics, I tried to filter out noise and identify what might actually hold up against elite competition. The data is the same as last week. The margin for error is not.

Model

Nothing fancy this time. No new tricks, no reinvention. The model is the same one that powered last week’s picks: logistic regression. Clean, restrained, and honest about what the data can and cannot say. With such a tiny playoff sample, that restraint matters. This model does not try to outsmart the chaos. It simply weighs the inputs, assigns probabilities, and moves on.

The results reflect that reality. In 2021, the model separated winners from losers with an AUC (definition in the first blog) of 0.8750. Strong, confident, and comfortably above randomness. In 2022, it cooled to 0.7500, still solid in a playoff environment where margins evaporate quickly. Then came the wobble. AUC dipped to 0.6250 in 2023, followed by a rough 0.2500 in 2024, the kind of number that reminds you how brutally unforgiving single-elimination football can be.

Averaged out, the model lands at 0.6250 AUC with a noticeably wider spread than last week. Translation: there is a signal, but it is faint, fragile, and wildly sensitive to context. This is not a crystal ball. It is a compass that sometimes spins when the wind picks up. And yet, in a sport defined by one bad drive or one miraculous comeback, even a modest edge is worth chasing. Same model, higher stakes, less room for error.

Results

Once the divisional round model locked in, the story sharpened fast. This stage of the playoffs does not reward volume or flash. It rewards precision. When the field narrows and everyone is good, the numbers that matter are the ones that show who can stay clean, take points when they are offered, and survive when the game tilts.

Offensively, efficiency ruled everything. Yards per play emerged as one of the strongest signals in the entire model at +0.53. Not total yards, not play count, just how much damage you do every snap. That pairs perfectly with Field Goal percentage, which actually carried the single largest offensive weight at +0.56. In the divisional round, drives do not need to end in touchdowns to matter. They just need to end in points. 

Defense made its presence felt in quieter but just as ruthless ways. Passer Rating allowed checked in at +0.36, a clear indicator that limiting quarterback efficiency still swings playoff games. Sack percentage sat even higher at +0.41, highlighting how much pressure matters when elite offenses are forced into obvious passing situations. 

Hovering over everything is Strength of Schedule at +0.45. Teams that had already been punched in the mouth during the regular season were simply better prepared for this environment. The divisional round has no warm-ups and no soft landings. Battle-tested teams tend to recognize moments faster, adjust quicker, and stay composed when the game tightens.

The message is clear. It is not about who looks dominant on paper. It is about who plays efficient offense, forces uncomfortable situations on defense, and has already proven they can survive against real competition. Everything else fades once the ball is kicked.

Divisional Round Predictions

Reasons

Buffalo Bills: Buffalo wins if Josh Allen turns the game into a stress test. Denver’s defense has shown cracks against good quarterbacks this season, and arguably, no one is better than Josh Allen. On the other side, if Denver’s rushing attack cannot exploit Buffalo’s biggest weakness, it puts Bo Nix in obvious passing situations against one of the league’s most disciplined pass defenses, something he has rarely been asked to do. That is a bad place to live in January.

Houston Texans: Houston’s path is built in the trenches. The Texans’ defensive front can overwhelm Drake Maye, who already took five sacks against a far less intimidating Chargers line. A turnover early, and the Texans can easily take control of this game. C.J. Stroud does not need to be spectacular here, just hold the ship. A clean game from him, supplemented by a good rushing performance against a good, not great run defense, will see the Texans advance past this round for the first time in franchise history.

Los Angeles Rams: This is a mismatch that jumps off the screen. The Rams bring the most complete offense in football into a game against a defense that simply cannot keep up. Chicago allows the 21st most passing yards in the league, and that is an open invitation for an offensive clinic. Yes, the weather will play a factor, but even if this turns into a ball-control game, the Rams have one of the best rushing attacks against a Chicago run defense that's worse than its pass defense.

San Francisco 49ers: Trent Williams and Sam Darnold. The 49ers need to give Brock Purdy enough time to dissect one of the league's best defenses, and the 49ers need Darnold to see ghosts. There is simply no other path to victory. Seattle's defense is good enough to carry them to victory even if Darnold has a bad game. For the 49ers to prevail, Darnold has to be horrendous, and Purdy has to have an awesome game.

There it is! My divisional-round predictions of the NFL playoffs. 1 favorite and 3 underdogs. Who do you think makes it out of this round? What's your prediction for the best game in this round?